1
00:00:00,000 --> 00:00:02,370
Welcome back. In this course,

2
00:00:02,370 --> 00:00:04,740
you'll learn how to use
TensorFlow to process

3
00:00:04,740 --> 00:00:07,410
texts for natural
language processing.

4
00:00:07,410 --> 00:00:09,180
Unlike images, which come in

5
00:00:09,180 --> 00:00:13,335
these regular shaped tensors
of pixel intensity values,

6
00:00:13,335 --> 00:00:15,629
Text is messier, there
are long sentences,

7
00:00:15,629 --> 00:00:16,980
there are short sentences.

8
00:00:16,980 --> 00:00:18,990
Do you want to process
individual characters

9
00:00:18,990 --> 00:00:20,860
alphabets or do you
wanna process words?

10
00:00:20,860 --> 00:00:22,020
So in this course,

11
00:00:22,020 --> 00:00:24,180
you learn how to
deal with all that.

12
00:00:24,180 --> 00:00:26,160
Yes. So like in our first week as

13
00:00:26,160 --> 00:00:28,260
well we're going to take
a look at what it takes

14
00:00:28,260 --> 00:00:30,150
for you to process text

15
00:00:30,150 --> 00:00:32,580
because neural networks
generally deal in numbers.

16
00:00:32,580 --> 00:00:34,110
The zero functions,

17
00:00:34,110 --> 00:00:36,240
calculating weights and
biases is all numbers.

18
00:00:36,240 --> 00:00:37,530
So how are we going to convert

19
00:00:37,530 --> 00:00:39,710
our texts into numbers
in a sensible way?

20
00:00:39,710 --> 00:00:41,425
Given a word like cat,

21
00:00:41,425 --> 00:00:42,750
how do you turn
that into a set of

22
00:00:42,750 --> 00:00:44,825
numbers that you can feed
into neural network?

23
00:00:44,825 --> 00:00:46,370
Exactly. Then if I have cat,

24
00:00:46,370 --> 00:00:47,855
then what's dog
going to look like?

25
00:00:47,855 --> 00:00:49,220
All of those kind of things.

26
00:00:49,220 --> 00:00:51,320
Then multiple length
sentences as well.

27
00:00:51,320 --> 00:00:52,970
How do we deal with padding them?

28
00:00:52,970 --> 00:00:54,260
Or if you've got like

29
00:00:54,260 --> 00:00:56,360
a body of words that
you use for training,

30
00:00:56,360 --> 00:00:57,800
and then you have another body of

31
00:00:57,800 --> 00:00:59,810
words that you want to
actually predict on,

32
00:00:59,810 --> 00:01:00,950
but you're going to
have some words in

33
00:01:00,950 --> 00:01:02,310
this body that
aren't in that one.

34
00:01:02,310 --> 00:01:04,430
How do you deal with out
of vocabulary tokens

35
00:01:04,430 --> 00:01:05,720
and that kind of things?

36
00:01:05,720 --> 00:01:06,660
There's going to be a lot of fun.

37
00:01:06,660 --> 00:01:08,550
Yeah. So in this first week,

38
00:01:08,550 --> 00:01:10,590
you learn how to
load in the texts,

39
00:01:10,590 --> 00:01:12,660
pre-process it and
set up your data

40
00:01:12,660 --> 00:01:15,300
so it can be fed to
a neural network.

41
00:01:15,300 --> 00:01:17,070
I'm really excited.
In this course,

42
00:01:17,070 --> 00:01:19,820
you'll learn how to deal
with texts using TensorFlow.

43
00:01:19,820 --> 00:01:23,220
So let's go on to
the next video to get started.